We've just switched to using the CLI upgrade script, and performed our first live upgrade using it from 3.4.2 > 3.4.4.
We are using memcached, and following the upgrade, cache/admin.php showed that memcached_cluster_store had 25 store mappings, while before the upgrade it had 39. The 14 missing cache definitions had no mapping set.
We have seen this issue before while using the web interface to run the upgrade, but only when it has timed out part-way through and we have had to re-run the script, never when it has run successfully first time. In this case, the CLI script ran without an error.
We are storing our dataroot on an NFS volume, which might be relevant here.
This seems like it's a bug in the upgrade process. Can anyone suggest where I might look to diagnose the issue further?